[infinispan-issues] [JBoss JIRA] (ISPN-6910) Random failures due to MagicKey

[Dan Berindei (JIRA)]

    [ https://issues.jboss.org/browse/ISPN-6910?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13278426#comment-13278426 ] 

Dan Berindei edited comment on ISPN-6910 at 8/12/16 8:07 AM:
-------------------------------------------------------------

[~rvansa] I guess it depends on what you expect {{MagicKey}} to do... if you want to generate a few keys per test, and you only had a low number of nodes, the chances of failing to generate a key are miniscule. I agree that the error message is pretty bad, but almost every time when 1000 iterations are not enough, it means there's no segment with the combination of owners you want.

If, however, you want to generate >10 keys, then I'd argue you don't really want {{MagicKey}}. You'd be much better served by writing a custom {{ConsistentHashFactory}} and/or {{KeyPartitioner}} that put every key in a particular segment, and assign that segment to the intended owners.


was (Author: dan.berindei):
[~rvansa] I guess it depends on what you expect {{MagicKey}} to do... if you want to generate a few keys per test, and you only had a low number of nodes, the chances of failing to generate a key are miniscule. I agree that the error message is pretty bad, but almost every when 1000 iterations are not enough, it means there's no segment with the combination of owners you want.

If, however, you want to generate >10 keys, then I'd argue you don't really want {{MagicKey}}. You'd be much better served by writing a custom {{ConsistentHashFactory}} and/or {{KeyPartitioner}} that put every key in a particular segment, and assign that segment to the intended owners.

> Random failures due to MagicKey
> -------------------------------
>
>                 Key: ISPN-6910
>                 URL: https://issues.jboss.org/browse/ISPN-6910
>             Project: Infinispan
>          Issue Type: Enhancement
>          Components: Test Suite - Core
>    Affects Versions: 9.0.0.Alpha3
>            Reporter: Radim Vansa
>            Assignee: Radim Vansa
>             Fix For: 9.0.0.Beta1, 9.0.0.Final
>
>
> There is a fixed 1000-iteration loop to find out proper key. In tests that create many magic keys, the cumulative probability of failing one of the keys' generation is going up and this is causing random failures in the testsuite.
> Moreover, the equality of magic keys is based only on the hashCode - with many keys, there is a non-trivial chance that the hashcodes will make two logically different keys equal.



--
This message was sent by Atlassian JIRA
(v6.4.11#64026)